28 Oct 2014

Is the Met Office’s new supercomputer worth £97m?

Today the Met Office announced that it is going to spend £97m on a new supercomputer, which will be one of the fastest computers in the world.

It marks an exciting point in weather and climate science, offering a greater amount of computing power with which to make predictions for the future.

However, I’m sure there’ll be a few of you who have rolled your eyes and questioned whether not this is worth it, given that, occasionally, the forecast still doesn’t quite go to plan.

What will it mean for weather and climate?

Weather computer models work by dividing our atmosphere into grid boxes, within which complex calculations are done to determine weather conditions at each point around the earth.

In simple terms, these calculations concentrate on the state of three fundamental elements that influence the weather on our planet: pressure, temperature and humidity.

Once these calculations have been made, a forecast can be produced that gives the weather for an area or location days into the future.

This new supercomputer will allow our atmosphere to be divided into even more boxes, leading to an increase in resolution, offering a more detailed, localised picture as to how the weather is expected to behave.

When I started off in weather forecasting a decade ago, the resolution of the data for the UK was 12km by 12km.

Significant improvements mean that this resolution currently stands at 4km by 4km, with 1km by 1km in the pipeline – something that this new supercomputer will make a reality.

It will not only be able to offer more detail, but offer the potential to look further into the future to hopefully increase the accuracy of weather predictions for weeks, even months ahead.

£97m is a lot of money – is it worth it?

While there are some who may say that £97m seems like a lot of money for a supercomputer, it’s important to look at the bigger picture.

You only have to look at the cost of the seemingly increasing number of severe weather events around the world, to see why this investment will more than pay for itself over time.

Great Storm Of 1987

The Great Storm of 1987 (see image above) cost an estimated £4.9bn in today’s money, for which a supercomputer of today’s capability would no doubt have provided a much better forecast and warning of what was to come.

Even though the storm would have happened anyway, greater warning of what was on the way would have probably saved lives and allowed government agencies to be more prepared.

It’s also important to remember that the Met Office provides forecasts for the whole world, as well as here in the UK – offering guidance for hurricanes and typhoons that affect millions of people.

Hurricane Sandy that slammed into the east coast of the US in 2012 was estimated to have been the second-costliest hurricane in US history, causing around $70bn of damage and directly responsible for the deaths of 148 people.

For me, these spiralling costs of severe weather events around the world, both financial and human, more than make the case for spending money on new supercomputers.

Continuing improvements to the quality and level of detail offered, will give even more warning in a world where extreme weather and a changing climate provides an ongoing threat.

Follow @liamdutton on Twitter 

Tweets by @liamdutton

6 reader comments

  1. Martin Pipe says:

    The prospect of more accurate weather forecasting? That’s only to be welcomed. However, my beef is that it’s £97m that private investors won’t have to stump up if (when?) the Tories’ ill-advised and ideologically-motivated privatisation of the Met Office goes ahead. Funnily enough the local branch of the Post Office here in Southend recently underwent an expensive refit; I doubt it was the only one in the country to undergo such refurbishment. And, by amazing coincidence, the Tories want to cast that public service to ‘the markets’ too. There are numerous other examples. But that’s Tory policy; I’m afraid. Get taxpayers to pick up the tab – its wealthy supporters then reap the rewards. And you thought that the whole point of privatisation was that it would save (rather than swallow) public money? No. It’s to make talentless crony-capitalists piles of effortless cash – at our expense. And you don’t need a supercomputer to work that one out…

    1. LB says:

      My concern is the electricity this machine will demand as it will be 10s of MegaWatts and the running costs have not been given. We have Electricity supply concerns right now and this will take a huge amount off the Grid. The other concern I have is that the modelling software used (UM & 4DVAR) are decades old and need to be re-written to make them more efficient and reduce the need for a power hungry Supercomputer. The tax payer needs to know this.

  2. Justin says:

    At least this article was better than the appalling Channel 4 news report I just saw which involved asking people in a park if they would like their weather reports read by a computer instead of a person.

  3. Caddy says:

    Best of luck with it, obviously, but I really do wonder if the only outcome will be that you’ll be able to get the wrong answer faster.

  4. Jon Wood says:

    I agree in principle that a finer mesh should lead to more accurate predictions, however weather is a chaotic system and predictions are vulnerable to measurement errors. I won’t be able to find a reference for this, but I am sure that the October 1987 weather data was re-run on a machine with a coarser mesh and correctly predicted Hurricane Fish – unlike poor Michael.

  5. Harry says:

    I’m amazed. 1km x 1km accuracy? It stretches credulity to the limits but we shall see. You’re not as specific about the accuracy of long range forecasting. Will we be able to book a weekend in Blackpool at the end of March with confidence?

Comments are closed.